Crescendo/Diminuendo Asymmetries in Beethoven's Piano Sonatas

Author(s): David Huron

Source: Music Perception: An Interdisciplinary Journal, Vol. 7, No. 4 (Summer, 1990), pp. 395-
402
Published by: University of California Press
Stable URL: http://www.jstor.org/stable/40285475
Accessed: 06/10/2010 17:52
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ucal.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
University of California Press is collaborating with JSTOR to digitize, preserve and extend access to Music
Perception: An Interdisciplinary Journal.
http://www.jstor.org
Music
Perception 1 990 by the regents of the
Summer
1990,
Vol.
7,
No.
4,
395-402
university of California
Crescendo/Diminuendo
Asymmetries
in
Beethoven's Piano Sonatas
DAVID HURON
Conrad Grebel
College, University of
Waterloo
A
study
of 32
piano
sonatas
by
Beethoven reveals a
significant asymme-
try
between
increasing
and
decreasing dynamics. Specifically: (1)
cre-
scendos are more
frequent
than
diminuendos, (2)
crescendos tend to last
longer
than
diminuendos, (3) large changes
of
dynamics
tend to involve
reductions in
loudness,
and
(4)
crescendos will more
commonly
follow
low
dynamic
levels than will diminuendos follow
high dynamic
levels.
These results
support
a
"ramp archetype"
of musical
dynamics
in which
the music tends to build in a
gradual way,
but tends to subside
relatively
quickly.
The results
parallel
a
previous study showing
an identical tex-
tural
asymmetry
in the evolution of
polyphonic
sonorities.
a
study
of voice entries and exits in multivoiced
music,
Huron
(1990)
demonstrated a
significant asymmetry
between
increasing
and decreas-
ing
textural densities.
Specifically,
it was found that voice entries tend to be
incremental
(one
at a
time)
whereas voice retirements tend to be
multiple
(several
at a
time). Moreover,
this same
pattern
was evident in the evolution
of the
density
of textural sonorities. This
pattern
of textural
change
can be
schematically represented
as in
Figure
1. In Huron
(1990),
it was found that
this
"ramp" paradigm
was less characteristic of
homophonic
music than of
polyphonic
music
-
at least as measured
by
the introduction and retirement
of
parts
or
by
the
density
of successive sonorities.
However,
it is
possible
that a similar
pattern
of textural
change may
be
present
in
homophonic
mu-
sic via some other musical
parameter.
One of the characteristic differences
between
baroque
and classical/romantic music is the latter's
emphasis
on
dynamic shading.
This
suggests
that an
appropriate way
to test for the
pres-
ence of the
"ramp archetype"
in
homophonic
music is to examine the do-
main of
dynamic
level.
In the realm of musical
dynamics,
the
ramp pattern suggests
the follow-
ing hypothesis.
Requests
for
reprints may
be addressed to David
Huron,
Music
Department,
Conrad
Grebel
College, University
of
Waterloo, Waterloo,
Ontario Canada N2L 3G6.
395
396 David Huron
Fig.
1.
Ramp pattern
of textural
density.
Hypothesis
There exists an association between the direction of
change
of
dynamic
level and the
magnitude
of the
change. Specifically, large (abrupt) changes
of
dynamics
will tend to involve diminutions of loudness while small
(grad-
ual) changes
of
dynamics
will tend to involve increases of loudness.
In this brief
paper,
we test this
hypothesis through
an examination of the
dynamic markings
found in the 102 movements of the 32 sonatas for solo
piano by Ludwig
van Beethoven.
Measuring Dynamics
One
way
to measure
dynamic changes
is to use a sound-level meter in
conjunction
with a chart recorder
(or computer equivalent)
to monitor the
sound-pressure
level of actual
performances.
This method will
generate
a
great
deal of data
-
but much of it will be difficult to
interpret
in
light
of the
above
hypothesis. Specifically,
note
onsets,
metric
accents,
and other fea-
tures are
apt
to confound the identification of increases and decreases of
dynamic
level. Such local fluctuations of level can be removed
by averaging
the
resulting data,
but this will have the undesirable effect of
smoothing
out
abrupt changes.
A much
simpler
method of
measuring
musical
dynamics
is to examine
the written
dynamic markings
found in a musical score.
Dynamic markings
themselves do not
necessarily
reflect the true
changes
in loudness over the
course of a work. Different
performers may
make
idiosyncratic
modifica-
tions
according
to their own
interpretive notions;
moreover these modifica-
tions
may
even contradict those found in the notated music. Other
changes
of
dynamics
will be
implied by
the score but not
explicit
in the written
Crescendo/ Diminuendo
Asymmetries
397
markings.
For
example,
a
passage may gradually
increase in
pitch height
-
suggesting
a crescendo that
may
not otherwise be
explicitly
identified in the
score. Nakamura
(1982, 1987)
has found that a
general
rise in
pitch
tends
to enhance the
impression
of a crescendo.
Notwithstanding
these
difficulties,
we
may
be confident that notated
dy-
namic
markings
hold some
degree
of
perceptual validity and, indeed, may
be
preferable
to
physical
measures of
sound-pressure
level
(Balzano, 1987).
In
particular,
where
dynamic markings
are notated with
great frequency
and
density,
we
may
be more assured that the written
markings
correlate to
a
large degree
with the
perceptual experience
of the music.
Beethoven's 32 sonatas for solo
piano provide
a
prime example
of a mu-
sical
repertoire copiously
annotated with
dynamic markings.
The Kalmus
edition contains
11,508 dynamic markings
in the 102 movements with a
mean
density
of notated
dynamics
of 0.49
per
bar. That
is,
on
average,
Beethoven nottes a
dynamic marking every
two bars. This
high
notational
density
makes Beethoven's music
especially
suitable for a
study
of
dynamic
changes.
Some of Beethoven's
markings
have a
questionable
value in terms of the
present study.
The
repertoire
contains some 3514 "sforzando"
markings
-
markings
that
may
be considered localized accents. In most cases it is best to
consider sforzando
markings
as
superimposed
on a base-line
dynamic
level.
But sforzando
markings
themselves
may
have subtle effects on the overall
level. For
example,
a
persistent
series of sforzandos
may imply
or be inter-
preted
as a crescendo that is not otherwise
explicitly
notated. It is thus dif-
ficult to resolve whether sforzando
markings ought
to be excised or re-
tained in the data
analysis.
In the
analyses given
below,
where successions
of
dynamic
levels are
examined,
two sets of results will be
given according
to whether sforzandos are eliminated or retained. In
general,
the results
turn out to be similar.
Crescendo/Diminuendo Measures: Method 1
Indications of
dynamic
level can be
arranged
as on the
following
scale
from loud to soft:
fff ff f mf mp p pp ppp
The
ramp hypothesis predicts
an association between the
magnitude
of a
change
of
dynamics
and the direction of that
change.
Successive
dynamic
markings
that are
neighbors
in the above list
may
be deemed
gradual
changes
of
dynamics. By contrast, abrupt changes
of
dynamics might
be
defined as
"jumps"
of two or more values
along
this
scale,
such as the
change
from forte
(/)
to
pianissimo {pp).
Tables 1A and IB
give
contin-
398 David Huron
TABLE 1
Dynamic Changes
A. Sforzando-Sensitive
Change
Crescendo Diminuendo
Gradual 153 184
Abrupt
291 547
B. Sforzando-Insensitive
Change
Crescendo Diminuendo
Gradual 218 205
Abrupt
479 751
gency
tables
relating abrupt
and
gradual dynamic changes
to the crescendo
and diminuendo conditions. Table 1A
presents
data where sforzando
markings
are deemed to
disrupt
transitions in
dynamic
level. Thus the se-
quence "piano"
followed
by
"sforzando" followed
by
"forte" is not
equiv-
alent to the
sequence "piano"
followed
by
"forte." Table IB
presents
data
where sforzando
markings
are deemed to have no effect on the base-line
dynamic
level. In this latter
case,
all sforzando
markings
are
ignored
in the
data
analysis;
hence the
sequence "piano"
followed
by
"sforzando" fol-
lowed
by
"forte" is
equivalent
to
"piano"
followed
by
"forte." Table 1 A is
thus a
proper
subset of Table IB.
Both
analyses
show the
predicted
association between the direction and
magnitude
of
change
and are
highly
consistent with the
ramp hypothesis
(Table
1A:
chi-square
=
11.65;
df
=
1; p
=
.000642; phi
=
+0.100;
Ta-
ble IB:
chi-square
=
20.47;
df
=
1; p
=
.000006; phi
=
+0.111).
More-
over,
the results show that there is little difference between the more strin-
gent
measurement conditions of Table 1 A and the less
stringent
conditions
of Table IB. In both
cases,
a small but
highly significant asymmetry
exists
between the manner in which the
dynamic
level builds and the manner in
which it subsides.
The values mezzo-forte
(mf)
and
mezzo-piano (mp)
are
comparatively
rare in Beethoven.
Hence,
it is moot whether the
marking "piano"(p)
fol-
lowed
by "forte"(/) ought
to be
regarded
as a
"large" change
of
dynamic
level. In Tables 1 A anc(
IB,
the
markings p- /'and f-*p
were deemed to be
abrupt changes
of level. A
reanalysis
of the data was carried out in which
these transitions were deleted. The
ensuing
sforzando-sensitive results were
found to be
insignificant (chi-square
=
1.87;
df
=
1; p
=
.170999;
phi
=
+0.053)
-
although
the data remain skewed
according
to the
pre-
dicted association. The sforzando-insensitive results remained
highly signi-
Crescendo/ Diminuendo
Asymmetries
399
ficant
(chi-square
=
17.42;
df
=
1; p
=
.00003; phi
=
+0.142).
This lat-
ter result
implies
that the
ramp hypothesis
is
supported
whether or not the
difference between
"piano"
and "forte" is considered to be
"large."
How-
ever,
the results are most consistent with the view that
forte-piano changes
constitute a
"large" change
of
dynamic
level.
Crescendo/Diminuendo Measures: Method 2
A
consequence
of the
ramp hypothesis
is that small
changes
of
dynamics
ought
to tend to increase rather than diminish the loudness. There are sev-
eral
ways
of
measuring
small or
gradual changes
of
dynamics.
The most
obvious
way
to
change dynamics gradually
is to
specify overtly
a "cre-
scendo" or a "diminuendo." Such
dynamic changes may
be indicated in the
score
by writing
the terms
"crescendo," "cresc," "cres.," "diminuendo,"
"dim.," "dimin.," "decrescendo," "decres.," etc.,
as well as other terms
such as "morendo." Our
hypothesis predicts
that where these
markings
ex-
tend for a
significant duration,
the number of crescendos should exceed the
number of diminuendos. In the Beethoven
sonatas,
1263 crescendo mark-
ings
are to be
found,
while
only
360 diminuendos are indicated. The num-
ber of crescendos is
approximately
3.5 times the number of diminuendos.
This ratio differs
significantly
from an
expected
ratio of
roughly equivalent
numbers at better than the 0.001 level of confidence and
supports
the
ramp
hypothesis.
Aside from the ratios of crescendos to
diminuendos,
we must also be
satisfied that the durations of the crescendos extend for a
significant period
of time
-
rather than
merely being
short-lived. For written
dynamic terms,
the effective
length
is
occasionally
indicated
through
the addition of dotted
lines, (e.g.,
"dim
-
in
-
u
-
en
-
do" or "cresc.
").
The total num-
ber of diminuendo
markings carrying
dotted-line extensions is
50,
while the
number of crescendo
markings carrying
dotted-line extensions is 222
-
more than a fourfold difference. For convenience in tabulation we can
define a
"long"
crescendo or diminuendo as one whose dotted-line exten-
sion extends over four or more bars
(i.e.,
crosses three
barlines).
If we com-
pare
the number of
long
diminuendos
(11)
with the number of
long
crescen-
dos
(121)
we see a tenfold
predominance
of
long-term
crescendos over
long-term
diminuendos.1 Hence we can claim that crescendos are more fre-
quent
than diminuendos and that crescendos
generally
extend over a
greater length
of time than diminuendos. These results
again support
the
ramp hypothesis.
1. A
perhaps arbitrary
and
inconspicuous
but
revealing
indication of the
greater length
of crescendos is the
proportion
of abbreviated terms:
only
2.5% of diminuendos/
decrescendos are
spelled
out
fully,
whereas 3.1% of crescendos are
spelled
out
fully.
400 David Huron
Crescendo/Diminuendo Measures: Method 3
An alternative method for
indicating
crescendos or diminuendos is
through
horizontal
V-shaped
or
"hairpin" graphical markings.
The
ramp
hypothesis predicts
that where these
markings
extend for a
significant
dura-
tion,
the number of crescendos should exceed the number of diminuendos.
Without
considering
the
length
of these
graphic devices,
the number of
graphic-diminuendos
in fact exceeds the number of
graphic-crescendos
in
Beethoven's
piano
sonatas. This
represents
a reverse trend to that
expected.
However,
this
apparent anomaly
can be resolved
by examining
the dura-
tions of these
markings.
For the
graphic hairpin symbols,
the
physical length
of the
printed sym-
bol
provides
a convenient indication of its
operative length.
Table 2
pro-
vides a breakdown of
hairpin
crescendo/diminuendo
markings by length
criteria. A
graphic hairpin symbol
was deemed to be
"long"
if it crossed
over one or more barlines.
Table 2 reveals a
striking
contrast.
Although
there are more
hairpin
di-
minuendos than
hairpin crescendos,
there is a
strong
association between
length
and the direction of
change:
diminuendos are
significantly
shorter in
length (chi-square
=
81.17;
df
=
1; p
=
.000000; phi
=
+0.248728).
Nearly
90% of
hairpin
diminuendos are less than a bar in
length
and so
cannot
truly
be said to be
"gradual."
If we
compare only longer
crescendo
and diminuendo
hairpin markings,
the
proportion
of crescendos is more
than double the number of diminuendos.
In
short,
Beethoven
generally
uses written terms such as crescendo or de-
crescendo as the
preferred
means of
indicating
those
changes
of
dynamics
that extend over several bars.
Hairpin
crescendo and diminuendo
markings
are reserved for
shorter,
more
local, changes
of
dynamics.
The
typical
hair-
pin dynamic marking
has an effect over a duration of less than a bar. When
hairpin markings extending
across one or more barlines are con-
sidered, hairpin
crescendos are
significantly
more
prevalent
than
hairpin
di-
minuendos. Once
again
these results are consistent with the
ramp hypothe-
sis.
table 2
Lengths
of
Graphic
Crescendos and Diminuendos
Length Crescendo Diminuendo
Long3 168 78
Short
392 674
aOne or more barlines crossed
by
the
graphic symbol.
Crescendo/ Diminuendo
Asymmetries
401
Crescendo/Diminuendo Measures: Method 4
A further test of the
ramp hypothesis
can be
pursued by examining
the
contextual
relationship
between
markings
of static level and
immediately
ensuing
crescendo or diminuendo
markings.
In this
case,
crescendos and
diminuendos
may
be indicated either
through
a written musical term or via
the notation of a
hairpin graphic symbol.
The null
hypothesis
would
posit
that a crescendo would be as
likely
to follow
p
or
pp
as would a diminuendo
to follow
for ff. However,
Tables 3 A and 3B refute this. If the
dynamic
level
is
high,
the
probability
of an
ensuing
diminuendo is
only
between 8 and
11%.2
However,
if the
dynamic
level is
low,
the
probability
of the subse-
quent marking indicating
a crescendo is over 40%. Once
again
this result is
consistent with the
ramp hypothesis.
Conclusions
A number of conclusions can be made
concerning changes
of
dynamics
in Beethoven's 32
piano
sonatas. The musical term crescendo is
signi-
ficantly
more common than the terms
indicating
diminuendo. When these
table 3
Context-Determined Crescendos and Diminuendos
A. Sforzando-Sensitive
Crescendo Condition Count Diminuendo Condition Count
PP-< 213/656 FF-> 33/651*
P -< 861/2027 F -> 67/1082
1074/2683 100/1733
interpret
as: 33 of 651 fortissimo
markings
are followed
by
the indication of a diminu-
endo.
B. Sforzando-Insensitive
Crescendo Condition Count Diminuendo Condition Count
PP-< 222/656 FF-> 61/651
P -< 935/2027 F -> 133/1082
1157/2683 194/1733
2. The
precise figure depends
on sforzando-sensitive or sforzando-insensitive contexts.
402 David Huron
terms are
accompanied by
dotted-line extensions
indicating
the duration of
the
dynamic change,
crescendos are
significantly longer
than diminuendos.
Similarly, graphically
indicated
"hairpin"
crescendos are
significantly
longer
than
corresponding
diminuendos.
Furthermore, large (abrupt)
changes
in
dynamics
are more
prone
to be reductions of loudness than in-
creases of loudness. In
addition,
crescendos will more
commonly
follow
low
dynamic
levels than will diminuendos follow
high dynamic
levels.
Together,
the above results
strongly
affirm the existence of a
ramp
arche-
type
in the musical
dynamics
of Beethoven's
piano
sonatas. In
general,
it
can be said that the music tends to build in a
gradual way,
but tends to sub-
side
relatively quickly.
This result
parallels
Huron
(1990),
where a similar
ramp archetype
was found in the introduction and retirement of voices in
polyphonic
music. In a
subsequent paper,
we will
develop
a
psychophysical
theory
as to
why
such an
asymmetry might
be favored in the
organization
of
music.3
3. The author extends his thanks to Dr. Mark
Haggard
of the British Institute of
Hearing
Research
(Nottingham, U.K.)
and to Dr. Robert Pascall of the
University
of
Nottingham
School of Music for
offering
valuable comments on an earlier draft of this
paper.
This re-
search was
supported
in
part through
funds
provided by
the Social Sciences and Humanities
Research Council of
Canada,
the Canadian Centennial
Scholarship
Fund
(U.K.),
and the
U.K. Committee of Vice-Chancellors and
Principals.
References
Balzano,
G.
Measuring
music. In A. Gabrielsson
(Ed.),
Action and
perception
in
rhythm
and
music. Stockholm:
Royal
Swedish
Academy
of
Music, 1987, pp.
177-199.
Beethoven, Ludwig
van. Sonatas
for
solo
piano.
From
Ludwig
van
Beethoven; Complete
edition
of
all his
works,
Vols. 20-22. New York: Edwin F.
Kalmus,
n.d.
Huron,
D. Increment/decrement
asymmetries
in
polyphonic
sonorities. Music
Perception,
1990, 7(4),
385-394.
Nakamura,
T.
[The
effect of
pitch pattern
on the
recognized
crescendo and
decrescendo] (in
Japanese). Reports of
the 1982
Spring Meeting of
the Acoustical
Society of Japan, 1982,
dd. 215-216.
Nakamura,
T. The communication of
dynamics
between musicians and listeners
through
musical
performance. Perception
and
Psych op hysics, 1987, 41(6),
525-533.
Patterson,
B. Musical
dynamics. Scientific American, 1974, 231(5),
78-95.